Predicting Recovery Boiler Performance and Exploiting Data Visualization for the Critical Variables

摘要

To analyze steam production, the techniques for variable selection via stepwise regression and genetic algorithms showed the best results as compared to the method using principal components. The variables selected for steam prediction by the neural networks were the outflow of the black liquor, the percentage of dry solids and the pressure in the steam drum. It is worthwhile to keep in mind that this result was obtained for operational data of the factory under study and that any result should be validated by specialists of the area, whether engineers or researchers. Due to the huge number of variables and of possible relationships among the variables of the processes, the techniques for variable selection and for data visualization are useful to help the operator make decisions, by selection of the relevant variables and allowing the work in a smaller space, and by allowing the visualization of the relationships among variables, respectively. The advantage is greater facility in manipulation as well as physical interpretation of the phenomena. This research will also be developed for the efficiency reduction and emissions.